The relationship between volatile compounds, metabolites and sensory attributes: A case study using lamb and sheep meat

(1)

Contents lists available atScienceDirect

Small Ruminant Research

journal homepage:www.elsevier.com/locate/smallrumres

The relationship between volatile compounds, metabolites and sensory attributes: A case study using lamb and sheep meat

Vladana Grabe ž

^a,

*, Milena Bjelanovi ć

^a

, Jens Rohlo ﬀ

^b

, Aleksandra Martinovi ć

^c

, Per Berg

^d

, Vladimir Tomovi ć

^e

, Biljana Rogi ć

^f

, Bjørg Egelandsdal

^a

aFaculty of Chemistry, Biotechnology and Food Science, Norwegian University of Life Sciences, P.O. Box 5003, NO-1432, Ås, Norway

bDepartment of Biology, Norwegian University of Science and Technology, NO-7491, Trondheim, Norway

cFaculty of Food Technology, Food Safety and Ecology, University of Donja Gorica, ME-81000, Podgorica, Montenegro

dNortura SA, Postbox 360 Økern, NO-0513, Oslo, Norway

eFaculty of Technology, University of Novi Sad, Bulevar cara Lazara 1, RS-21000, Novi Sad, Serbia

fAgricultural Faculty, University of Banja Luka, Bulevar vojvode Petra Bojovića 1A, BA-78 000, Banja Luka, Bosnia and Herzegovina

A R T I C L E I N F O

Keywords:

Sheep

Adipose tissue volatiles Lean meat metabolites Sensory attributes

A B S T R A C T

The aim of this study was to use aflavoromics approach to identify key compounds responsible for sensoryflavor of lamb and sheep meat. The investigation was confined to volatile compounds from adipose tissue and metabolites in lean meat using headspace-gas chromatography/mass spectrometry (HS-GC/MS) and solvent extraction-GC/MS, respectively. Partial least square regression analysis supported with variable selection were used to correlate identified compounds to sensory attributes. Several metabolites involved in energy production via Krebs cycle and Embden-Meyerhof-Parnas pathway contributed to gamy and grassflavor. Gamyflavor was strongly and positively correlated with aspartic acid, cyclo-leucine, gluconic, citric and pyruvic acid. Gluconic and pyruvic acid together with formic acid,β-caryophyllene, 3-methylphenol, 2-ethylfuran showed strong positive correlation with grassflavor. Sugars (glucose, mannose-6-phosphate and myo-inositol) were negatively correlated with gamy and grassflavor, suggesting a role in suppression of off-flavors in lamb and sheep meat.

Bitterflavor was strongly correlated with hypotaurine and (E)-2-pentenal. Metallicflavor and bitterness were influenced by almost the same compounds. Acidicflavor was not explained by any compound identified, while rancidity was not detected by panelists. Finally, theflavor components describing grass and bitterflavor could be used to discriminate animals from different production systems.

1. Introduction

Meatflavor is an important quality criterion with a key role in the overall lamb/sheep meat acceptability (Wood et al., 1999). Significant attention has been given to the characteristic mutton and pastoral flavor that negatively affects consumers’ acceptance of lamb/sheep products (Sink and Caporaso, 1977;Young et al., 2003). Muttonflavor is described byWong (1975)as sweaty, sour, urinary, fecal, barnyard, oily, sharp and acrid. Thisflavor note was associated with branched chain fatty acids (BCFA; C8 – C10), specifically 4-methyloctanoic, 4- ethyloctanoic and 4-methylnonanoic acids that are more abundant in adipose tissue of aged animals (Wong et al., 1975a,1975b; Watkins et al., 2013). However, it is noteworthy that discrimination of lamb from sheep meat according to BCFA concentration has never been reported as possible (Watkins et al., 2010).

Pastoralﬂavor described as sheepy, gamy, barnyard, animal, fecal, is related with pasture-fed animals (Schreurs et al., 2008). Thisﬂavor note has also been associated with higher concentrations of 3-methy- lindole (skatole) and 4-methylphenol in lamb adipose tissue (Young et al., 2003).

The chemistry offlavor is very complex and depends of interaction between volatile (aroma) and non-volatile (taste) compounds. A number of studies have been carried out to identify and define key volatile compounds associated with the characteristicflavor in cooked sheep meat (Almela et al., 2010;Bueno et al., 2014;Caporaso et al., 1977;Elmore et al., 2000;Hornstein and Crowe, 1963;Resconi et al., 2010;Young et al., 1997). Generally, limited work has been done on non-volatile (metabolites) compounds and their role in lamb/sheep flavor (Watkins et al., 2013). In addition, the complex nature of meat flavor requires understanding of the essentialflavor-active compounds

https://doi.org/10.1016/j.smallrumres.2019.09.022

Received 20 February 2018; Received in revised form 15 June 2019; Accepted 30 September 2019

⁎Corresponding author.

E-mail address:[email protected](V. Grabež).

Available online 12 October 2019

T

(2)

isolated both from adipose tissue and lean meat and their joint contribution to perceivedﬂavor.

To understandflavor properties of lamb/sheep meat, in the present study an untargeted approach called flavoromics was applied (Ronningen, 2016). This novel approach in flavor research combines three phases: characterization of volatile and non-volatile (metabolites) compounds, model development and validation of compounds. The analytical information, as an outcome of these three steps, is correlated with sensory properties in order to define compounds responsible for specific attributes. Using this approach, the aim was to: 1) Identify and quantify volatiles and metabolites as constituents of lamb/sheep meat flavor; 2) Evaluate sensory properties of lamb/sheep meat; 3) Elucidate how volatiles and metabolites from different metabolic pathways correlate with sensory attributes using aflavoromics approach.

2. Materials and methods 2.1. Experimental design

Ninety-two female animals were used in the study. In order to get high variability in ﬂavor proﬁles the following animals were chosen:

lambs (5−6 months), young sheep (∼2 years) and old sheep (4−5 years) belonging to two diﬀerent breed representative for the production system of three country of origin (Bosnia and Herzegovina–BH, Montenegro–MN, and Norway–NW). Lamb (18 animals; NW lamb), young (15 animals; NW 2y) and old sheep (14 animals, NW 4y) belonging to the Norwegian White Sheep breed were selected.

Furthermore, lamb (BH lamb) and old sheep (BH 4y), 15 animals each, belonged to Vlašićka Pramenka, being the most common phenotype of Pramenka breed in BH. Thus,ﬁfteen old sheep of Pivska Pramenka from Montenegro (MN 4y), as a second Pramenka phenotype, were included in this experiment. Six months old lambs of Pivska Pramenka from the same herd could not be obtained.

2.2. Tissue sampling

All animals were slaughtered in the country of origin (for more details seeBjelanovićet al., 2015). TheM. longissimus thoracis et lum- borum(LTL) from left side of carcass was removed and adipose tissue available on the surface of the muscle was excised within 24 h post mortem, wrapped in aluminum foil, vacuum-packed and stored at -80 °C. A slice of LTL was vacuum packed and stored at -80 °C for intramuscular fatty acid analysis. The rest of LTL was vacuum-packed, refrigerated for 7 days (at 4 °C), divided into slices of 2.5 cm thickness, vacuum-packed and stored at -80 °C for sensory and GC/MS analysis.

All samples were analyzed in the same laboratory.

2.3. Fatty acid composition

Intramuscular fat was extracted according to AOAC Oﬃcial Method (AOAC 991.36, 1996). Fatty acid methyl esters (FAME) synthesis was performed according to modiﬁed method byYi et al. (2013). The fatty acids were analyzed by accredited laboratory (http://vitas.no/) according to theO’Fallon method (2007).

2.4. Extraction, derivatization, and GC/MS analysis (GC/MSextraction) of meat metabolites

One gram of lean meat was transferred into a 15 mL tube, and 5 mL of a water: methanol: chloroform (1: 2.5: 1) mixture with internal standard ribitol (66μg/mL) was added. The sample was incubated at 60 °C for 60 min in sonication bath and centrifuged for 10 min at 3 000 rpm at 4 °C. An aliquot of 1 mL was transferred into a 1.5 mL Eppendorf tube, dried in a SpeedVac (Thermo Scientiﬁc, Waltham, MA, USA) overnight and stored at -80 °C. The dried residues were re- suspended in 80μL methoxyamine hydrochloride with pyridine

(20 mg/mL) at 30 °C for 60 min and sonicated at 30 °C for 30 min.

Finally, samples were treated with 80μL of N-methyl-N-(trimethylsilyl) triﬂuoroacetamide at 37 °C for 30 min.

GC/MS analyses were performed according toSissener et al. (2011).

Derivatized samples (1μL) were analyzed on an Agilent 6890 GC connected with an Agilent 5975 MS detector. A HP-5MS capillary column (i.d. 30 m ×0.25 mm,ﬁlm thickness 0.25μm) was used. The carrier gas (He)ﬂow rate through the column was 1 mL/min. The GC temperature program: 70 °C for 5 min, ramped at 5 °C/min until 310 °C. Analysis time was 60 min. The MS was operated at 230 °C, and the recorded mass range wasm/z50−700.

MS ﬁles from Agilent ChemStation (Agilent Technologies, Waldbronn, Germany) were exported in the netCDF format (OPENChrom, Eclipse Public License 1.0) to MetAlign (version 041012, RIKILT Wageningen UR, Plant Research International) for data pre- processing and alignment. Metabolites were identiﬁed with the AMDIS software (version 2.71, National Institute of Standards and Technology, Boulder, CO, USA) in combination with NIST05 (National Institute of Standards and Technology/Gaithersburg, MD, USA) and GOLM meta- bolome database (Max-Planck Institute for Molecular Plant Physiology, Golm, Germany). Normalization of the peak area was performed on the internal standard ribitol and expressed as mg/kg of meat. Samples were run randomized. Metabolites are presented in Table S-2.

2.5. Headspace gas chromatography/mass spectrometry (HS-GC/MS) analysis of volatile compounds

Frozen adipose tissue was homogenized with a crushing machine (IKA®A11 Basic Analytical Mill, Staufen, Germany) to aﬁne powder.

Four grams of homogeneous tissue were placed in a glass vial (50 mL) and stored at -80 °C until the next preparation step. In order to increase the volatile compounds extraction and generate representative volatile proﬁles, the homogenized sample was heated at 75 °C in water bath for 30 min on the day of analysis. This treatment improved extraction of volatile compounds from adipose tissue in agreement with Sivadier et al. (2008).

The liquid fat phase (1 g) was transferred to a clean glass vial and kept at 4 °C for∼4 h before measured. All samples were analyzed in two replicates.

A mixture ofﬁve compounds in Mygliol (AXO INDUSTRY, Warve, Belgium) was used as a control sample throughout the measurement period, at the beginning and end of sequences. These compounds were:

butanal (99%),cis-2-penten-1-ol (95%), 2-undecanone (99%), and dimethyl sulfone (98%) (Sigma-Aldrich Chemie GmbH, Schnelldorf, Germany) and acetic acid (100%, VWR, Fontenay-saus-Bois, France).

HS-GC/MS analysis was performed according to a method modiﬁed fromVolden et al. (2011). The extraction of volatile compounds from 1 g of liquid fat was performed on dynamic headspace analyzer Tele- dyne Tekmar HT3 (Teledyne Tekmar, Ohio, USA). The sample temperature (75℃) during headspace extraction step resulted in an un- satisfactory analytical signal. Therefore the temperature was increased to 150℃to improve the extraction and signal quality of volatiles from fat samples. This procedure may unintentionally introduce some reac- tion products due to heating in addition to the volatiles present at lower temperatures.

The compounds were analyzed by Agilent gas chromatograph 6890 N (Agilent Technologies, Santa Clara, CA, USA). The DB - WAXetr fused silica capillary column (30 m ×0.25 mm i.d., 0.50μm film thickness; J&W Scientific, USA) was connected to the ion source (230 °C) of a Agilent 5975 (Agilent Technologies, SantaClara, USA) quadrupole mass spectrometer (interface line 250 °C). The carrier gas was He with aflow rate of 1.0 mL/min. The temperature program for GC was: 35 °C for 10 min, ramped 1.5 °C/min up to 40 °C, ramped 4.0 °C/min up to 70 °C, ramped 7.5 °C/min up to 230 °C, and 1 min at 230 °C. Analysis time was 54.62 min, and recorded mass range wasm/z 33−300. Volatiles were identified by: (i) computer-matching of

(3)

generated mass spectra with NIST05 database (National Institute of Standards and Technology/Gaithersburg, MD, USA) and (ii) comparison of retention indices (RIs) with published RI values. Identified compounds (Table S-1) were used for statistics (see below). All compounds referred to below except butyrolactone that we failed to ac- quire, have been re-identified using pure compounds. The standard solutions run during measurement period were run at four different concentrations (R²= 0.996–0.999 for regression line). The concentration for all volatiles was standardized to the calibration curve for most relevant chemical compound present in the standard mix described above.

The two GC methods (GC/MSextractionand HS-GC/MS)were selected with priority on identiﬁcation of odor and taste related compounds in lamb/sheep.

2.6. Sensory analysis

For sensory testing meat samples were defrosted at 4 °C overnight.

The 2.5 cm slices of lean meat were heated in water bath set to 80 °C until internal temperature of 71 °C was achieved (AMSA, 1995) and served as 1 × 1 × 1 cm pieces to each assessor. A panel consisting of 8 trained (ISO 8586–1:1993) assessors (4 females and 4 males 30−59 years old) was selected for the sensory analysis. The laboratory for sensory analysis at Faculty of Technology in Novi Sad was designed according to ISO 8589:2007. During the evaluation, water and bread were served to assessors to cleanse their palate between samples. An- imal group was randomly selected, and then the whole group was analyzed. Three samples were served per session and two sessions were performed. Sensory traits of lamb/sheep meat were evaluated by the quantitative-descriptive analysis (Lawless and Heymann, 2010), using a scale from one (none) to nine (very intense) according to ISO 4121:2003. Assessors were asked to evaluate the following odor (gamy, grass, rancid) and taste (acidic, bitter, metallic) attributes. Gamy was deﬁned like leather/ horse saddle and grass odor like cut grass. Ferrous sulfate (FeSO4× 7H2O) was used as reference standard for training of assessors for sensory evaluation of metallic taste (ISO 3972-1991).

These attributes were selected as they have been observed to distin- guish diﬀerent Norwegian lamb samples earlier (Lind et al., 2011). The other tastes were deﬁned as in basic taste tests (see below). All samples were analyzed in the same sensory laboratory.

2.7. Flavor threshold determination

Flavor threshold values were obtained from VCF database (Volatile Compounds in Food 16.6.1;https://www.vcf-online.nl/VcfCompounds.

cfm) unless otherwise stated (see Appendix A. Supplementary Material - Table 1 and 2). Potentially interesting compounds had concentrations higher than reportedflavor thresholds in air or water/ fat/ oil. External standards from different chemical groups and applied at different concentrations were used for quantification. The analysis is semi- quantitative.

There were no literature data for theflavor thresholds of some relevant compounds dissolved in water. Flavor thresholds for six compounds (2-heptadecanone, gluconic acid, dimethyl sulfone, hypotaurine, mannose-6-phosphate and uridine) that correlated with sensory attributes (gamy, grass, bitter) were identified using 2-AFC method. In order to define minimum and maximum concentration of each compounds for threshold study, preliminary survey was performed based on Maximized Survey-derived Daily Intakes value (MSDI-EU;

http://www.thegoodscentscompany.com; Perfumer and Flavorist, 2017a, 2017b, 2017c).

The four basic tastes sweet (sucrose), salty (sodium chloride), sour (citric acid monohydrate) and bitter (caﬀeine) were prepared as solutions in deionized water and stored at 4 °C in screw glass bottles. Six sensory experienced persons (31–43 yrs old) were assembled at the Norwegian University of Life Sciences. Prior toﬂavor threshold analysis

panelists were re-trained by tasting easily recognizable solutions of sucrose (90 mM), sodium chloride (340 mM), citric acid monohydrate (3260μM) and caﬀeine (1.75 mM) as suggested byGomez et al. (2004) andTorrico et al. (2015). For“blanks”deionized water was used.

On the day of analysis, participants were invited at 11.00 o’clock and instructed to have a light breakfast and avoid smoking, drinking coffee, tea, refreshments or chewing gum for at least 2 h before the test (Gomez et al., 2004). Five solutions of each compound, from 4.7 to 75 mg/kg for 2-heptadecanone, gluconic acid and dimethyl sulfone, from 6.25 to 100 mg/kg for hypotaurine and mannose-6-phosphate, and from 0.625 to 10 mg/kg for uridine were prepared in 250 mL graduated closedflasks using deionized water and stored at 4 °C. The samples forflavor threshold analysis (10 mL) were presented in 15 mL plastic tubes labeled with a 3-digit random code.

For the series of trial, participants were presented withfive different concentrations of each chemical in order of increasing concentration until they report difference between chemical solution and distilled water. Upon comparing the samples, being two distilled water samples and one chemical solution sample, the subjects expressed freely their impressions aboutflavor profiles for chemical solutions using their own expressions. Subjects were informed about chemical safety information for all compounds and the purpose of the test. Threshold concentration was defined as a concentration of compound at which panelists could detect a difference from deionized water 50% of the time. Some compounds (see below) remained withoutflavor thresholds; one because no pure compound was available and the rest were excluded since they were described as hazardous or there were no available data about their toxic effect.

2.8. Statistical analysis

Sensory scores and fatty acid composition were analyzed using Microsoft Excel 2016, considering the animal group as a single unit. The sensory data have only been used for regression analysis where analysis of variance of sensory data were not relevant. In order to explore the relationship between chemical compounds and meat flavor, Partial least squares regression (PLS) and Principal component analysis (PCA) were carried out using Unscrambler, version X10.1 software (Camo, Trondheim, Norway). The PLS routine was used and the calculations were made in 3 manners. First, the data were kept as is, secondly only the volatiles were multiplied by 100 so that the dimension was changed to (μg/kg meat×10). The third data set was generated by multiplying only the volatiles by 1000 so that data originally calculated with dimension mg/kg now appeared with the dimensionμg/kg meat. This dimension weighting was done to keep the compounds at comparable magnitudes and thereby increase the possibility that a compound present in low quantities could be included in the explanatory model. In addition, the concentrations defining the 3 concentration weighted matrices above were used for modelling both with and without weighting according to standard deviation (SD) of a specific variable/

compound (mean/SD). The latter was an additional principle to chan- ging dimension (from mg toμg orμg×10) to secure that both compounds in low and large quantities could enter the explanatory models.

No compound was selected as influencing a sensory attributes unless the compound’s regression coefficientβ(w) significantly (P< 0.05) differed from zero (equal to zero defined no influence) for all 3 different weighting (×1, ×100, ×1000) principles. The above procedure was selected to reduce the prevalence of false positive associations. It should be pointed out that the procedure to a large extent eliminated the need for correct absolute concentrations. Only the flavor threshold are strictly dependent on absolute concentrations.

The PLS model was set up with random validation using segments of 7 samples; that is approximately 50% of an animal group. As an example, one group would be lamb from a speciﬁc region. It was, however, not critical for the results whether the segment number for validation was higher (e.g.10) or lower (e.g.4) than 7.

(4)

Principal component analysis (PCA) were performed to visualize flavor (local/global) markers significant for different animal groups in a reduced dimension plot. The PCA models included volatile compounds isolated from adipose tissue and metabolites from lean lamb and sheep meat that were significantly different (P< 0.001) between animal groups. Sample names were coded as described in the Experimental design section.

3. Results and discussion 3.1. Fatty acid composition

Fatty acid (FA) composition indirectly plays an important role in characteristic meatﬂavor in various animal species (Kosowska et al., 2017). Fatty acids are directly or indirectly involved in generation of the volatile compounds andﬂavor constitution (Van Ba et al., 2012). In our study, the total fatty acids of intramuscular fat ofM. longissimus thoracis et lumborumin 92 female animals was 50 mg/g of meat, with saturated fatty acids (SFA), monounsaturated (MUFA) and poly- unsaturated fatty acids (PUFA) presenting 50.5%, 42.1% and 7.4%, respectively (for more details see Bjelanović et al., 2015). PUFAs, namely, α-linolenic acid (2.0%), eicosapentaenoic acid (EPA; 0.4%), docosapentaenoic acid (DPA; 0.5%), docosahexaenoic acid (DHA;

0.2%) were considered. These FA may causeﬂavor defects as a result of the oxidation induced by cooking (Watkins et al., 2013). Higher levels of oxidation products were previously found for grilled meat from lambs fed supplement rich in EPA and DHA (Elmore et al., 2000,2005).

Campo et al. (2003)found that mixtures of α-linoleic acid, cysteine, ribose and iron were associated with‘grass’ﬂavor and related to meat from grazing animals.

3.2. Sensory attributes

The sensory attributes of all animal groups are presented inTable 1.

The assessors used the scale from 1−9 to evaluate odor and taste.

Gamy and grass odor were clearly identiﬁed, bitterness was less well identiﬁed, while metallic and acidic taste had limited variation for the examined samples. The assessors did not identify rancidity for the samples and the attribute was therefore excluded.

3.3. Correlation among volatile compounds, metabolites and sensory attributes

In the present study an untargeted approach was used to identify volatile compounds and metabolites in lamb and sheep adipose tissue and lean meat, respectively. Adipose tissues of 92 animals were analyzed. Seventy-five volatile compounds were identified and classified according to their chemical nature (Supplementary Material, Table S- 1). They were alkanes (15), alkenes (8), alcohols (11), aldehydes (19), ketones (7), acids (5), lactone, terpene, sulphur compound, phenol, ester and others. Among the selected compounds some were found in adipose tissue of one animal group but not in others, i.e. 3-methyl- hexane, 3-methylphenol. 3-methylphenol was identified only in adipose

tissue of MN 4y sheep. Approximately 50% of all identified volatile compounds were lipid oxidation products. The identified volatiles in our study were in agreement with previously reported volatile profiles of lamb fat and meat heated to lower (< 90℃) temperatures (Osorio et al., 2008;Sivadier et al., 2008;Vasta et al., 2007,2012). Few cyclic compounds were formed (e.g. cyclic alkenes) that may indicate stronger heat treatment. In addition, 69 metabolites were separated and identified in the lean meat using GC/MSextractionanalysis (Table S-2), although unlike volatiles these metabolites were identified in all animals but at different levels.

Interaction between odor (volatiles) and taste (metabolites) compounds is important for meatﬂavor although taste is dominating re- sponse. However, in further discussion for odor and taste attributes we will use termﬂavor since taste were assessed with open nostrils.

PLS models provided correlations between sensory attributes and all measured chemical compounds. The impact of key volatile and metabolite compounds on sensoryﬂavor properties was determined based on PLS analysis (P< 0.05) as variable selection criteria.

Only 17 volatile compounds and 19 metabolites significantly (P< 0.05) correlated with sensory attributes in PLS regression analysis (Fig. 1). Chemical compounds that did not pass the selection criteria are listed in Table S-3 (Supplementary Material). Compounds could be left out, i.e. not selected, because: 1) they do not correlate significantly to sensory attributes despite being present aboveflavor threshold; 2) their measured value varied too much for significance despite having a re- levantflavor. Twenty-six of the metabolites and 31 volatile compound were eliminated byfirst selection criteria. Metabolites presumed above sensory threshold, but left out were: malic acid (acid), sucrose (sweet), fructose (sweet), cysteine (sulpherous) and 4-aminobutyric acid (savory). Volatiles that were aboveflavor thresholds but left out were pentanal and octanal (fruity note), (Z)-4-heptanal (green), dimethyl sulphone (sulfurous/ metallic) and toluene (complex note). To some extent (Z)-4-heptanal (green), dimethyl sulphone (sulfurous/ metallic) may seem relevant for grass or metallicflavor but nevertheless these compounds clearly failed the selection criteria.

3.3.1. Gamyﬂavor

Gamyﬂavor was well explained (53-51%) and 15 compounds were included in the model (Fig. 1a). The explanation was highest when components present in low quantities were selected in all 3 models (×1, ×100, ×1000) by weighted variables (1/SD).

In principle, there could be 6 independent compounds or cluster of compounds that contributed to gamyflavor. This is because maximum 6 uncorrelated principal components (PC) were identified. Thus, it is possible that more components are associated with gamyflavor since compounds metabolically can correlate with other compounds. Krebs cycle components (e. g.citric acid, succinic acid) can be related since these are involved in respiration. In this way malic acid may indirectly contribute toflavor, albeit not being selected, since succinic acid af- fectedflavor. In addition, it is also important to note that the compounds that apparently suppress gamyflavor can also camouflage this sensory attribute and therefore indirectly affect theflavor, i.e. glucose can modify/reduce otherflavors (Meinert et al., 2009).

Amino acids reflect several physiological situations in the animal species: fatigue, stress, postprandial time etc., possibly reducingflavor acceptability (Warner et al., 2007) and alter the experience of a gamy flavor. Maruri and Larick (1992) reported a positive correlation between diterpenoids and off-flavor of grass-fed beef described by sensory panelists as gamy/stale. The actual gamyflavor is so far not well explained in terms of chemical compounds. Aspartic acid and cyclo-leucine showed positive, while glycine negative correlation with gamy flavor. Aspartic acid and glycine were identified as having concentrations aboveflavor thresholds, where glycine with its slightly sweet note (Drauz et al., 2007) may reduce gamyflavor. Cyclo-leucine is a non proteinogenic amino acid that may be a product of ruminal bacteria.

Cyclo-leucine can be related to the level of methionine that correlated Table 1

Sensory quality proﬁle (evaluated on a 1−9 scale) assessed by trained assessors on lean meat.

Sensory attributes Mean^a SD^b Min value Max value

gamy 4.4 1.0 2.6 6.5

grass 2.6 0.5 1.8 3.8

acidic 2.1 0.3 1.4 2.8

bitter 2.0 0.5 1.1 3.4

metallic 1.6 0.4 1.0 2.5

a Mean = average scores for each attribute for 92 animals.

b SD = standard deviation.

(5)

positively to gamyflavor but did not pass the selection criteria. In addition, cyclo-leucine correlated significantly (P< 0.001, linear regression) to aspartic acid, glycine, homocysteine, leucine, lysine, phenylalanine, tryptophan and tyrosine. Therefore, cyclo-leucine together with aspartic acid possibly presents a marker not only for gamyflavor but also for the general catabolic/anabolic status in lamb/sheep that may affectflavor.Nishimura et al. (1988)reported that higher content of free amino acids corresponded to higher intensity of brothy taste of pork and chicken meat. Our results suggest that bouillon-like taste of aspartic acid was probably reduced by bitter taste amino acids, i.e.

leucine, lysine, phenylalanine, tyrosine (Schlichtherle-Cerny and Grosch, 1998). The concentration of 4-hydroxyproline, as sweet taste metabolite (Wieser et al., 1977), was belowﬂavor threshold (Fig. 1a) but had a positive correlation to compounds like hypotaurine (r = 0.60, P< 0.001), gluconic acid (r = 0.68,P< 0.001), ascorbic acid (r = 0.55, P< 0.001), uridine (r = 0.61,P< 0.001) and hexanal (r = 0.57,P< 0.001). In addition, 4-hydroxyproline showed positive correlation to only one amino acid,β-alanine (r = 0.51,P< 0.001) but this amino acid is not important in protein biosynthesis. This cluster of compounds is discussed below.

Three organic acids (citric acid, pyruvic acid and succinic acid) were correlated with gamyflavor, where citric and pyruvic acid were positively correlated. Pyruvic acid was belowflavor threshold, but correlated to several compounds in the citric acid cycle including a positive correlation to malic acid (r = 0.30, P= 0.004).Lugaz et al. (2005) reported that both citric and malic acids evoke stronger sour sensation than lactic or acetic acid at equal concentration. In addition, pyruvic acid correlated to sucrose (r = 0.61; related to pyruvate production through glycolysis,P< 0.001) that was present in concentration above flavor threshold. The sugar acids (gluconic, glyceric and ribonic acid) also correlated to pyruvic acid suggesting that the metabolic status of

sugar polymer degradation may be involved in deﬁning gamyﬂavor.

Succinic acid, as metabolite that contributes to umami taste of Swiss cheese (Cadwallader and Singh, 2009) and chicken broth (Dunkel and Hofmann, 2009), showed negative correlation (r = 0.4, P< 0.001) with gamyﬂavor. Pyruvic acid, as a key compound in several metabolic pathways, and other Krebs cycle substrates are likely involved in the development of gamy ﬂavor as many wild animals have oxidative muscles (Curry et al., 2012) that need to be furnished by the Krebs cycle to produce ATP for extensive movements.

Glucose and mannose-6-phosphate can be regarded as one variable due to their correlation (r = 0.57,P< 0.001). Glucose was present in a far higher concentration than the other metabolites and aboveflavor threshold, possibly it reduced the intensity of gamyflavor. Mannose-6- phosphate positively correlated to two metabolites from Embden- Meyerhof-Parnas pathway, fructose-6-phosphate and glucose-6-phosphate (r = 0.80 and r = 0.90, respectively,P< 0.001), but also to many other components like fructose. Strangely enough sucrose tended to enhance gamyflavor (r = 0.39,P< 0.001). Myo-inositol is a car- bohydrate/sugar alcohol with half the sweetness of sucrose and it was negatively correlated with gamy flavor.Koutsidis et al. (2008) suggested that higher concentration of sugars and sugar-related compounds in beef meat is related with higher glycogen levelpre-mortem and/or intensified glycolytic activitypost-mortem.

The volatiles 2-methylheptane, 3-methylheptane, octane and (E)-2 octene can actually be looked upon as one variable due to positive correlation between the 4 selected compounds (r = 0.89 – 0.94, P< 0.001,Fig. 1a). Only (E)-2-octene may be above threshold since the other volatiles were below. However, (E)-2-octene is classified as dangerous (European Chemicals Agency - ECHA) organic compound and its threshold was therefore not determined. In addition, this volatile compound may not be a universal marker for gamy flavor since the Fig. 1.Estimated regression coefficients (βw; dimension (1/(mg/kg) for VOC and metabolites) obtained from Partial least squares regression (PLS) analysis of chemical compounds (X) and sensory attributes (Y). Beta (w) coefficients were used for relation between sensory attributes and volatiles and between sensory attributes and metabolites, respectively. Black bars in a plot represent compounds with concentrations higher thanflavor threshold (odor, taste or both), grey bars are compounds with concentrations lower thanflavor threshold, and bars with pattern are compounds with unknownflavor thresholds.

(6)

compound was not identiﬁed in all animals, but was most typical for sheep from Montenegro. (E)-2-octene has previously been detected in beef fat obtained from the renal periphery of beef carcasses (Umano and Shibamoto, 1987), in grilled lamb meat (Madruga et al., 2013) but also in fermented meat (Hui and Evranuz, 2012). The selected volatiles (2- methylheptane, 3-methylheptane and octane) had a positive correlation to citric and pyruvic acid (r = 0.67–0.78,P< 0.001) in addition to several other lipid degradation products for example octane (r = 0.88, P< 0.001).

Metabolites that correlated with gamyflavor were identified in all animals at various levels (Table S-2) and can therefore be universal markers of gamyflavor, while important volatiles were commonly only found in a fraction of the samples and cannot be equally suitable markers.

3.3.2. Grassﬂavor

As for gamyﬂavor, maximum 6 compounds or clusters of correlated compounds were suggested for grass ﬂavor by the PLS analysis.

Nineteen compounds, volatiles and metabolites (Fig. 1b), explained grassflavor up to 50% depending on weighting. Grassflavor positively correlated (r = 0.62, P< 0.001) with previously described gamy flavor. Therefore it is expected that some compounds that contributed to gamyflavor will also be relevant for grassflavor. It should be noted here that no amino acid was positively correlated to grassflavor or above its threshold. However, 4-hydroxyproline was included as relevant based on its correlation to other compounds.

Only two acids, gluconic and pyruvic acid, were positively correlated to grassflavor. Gluconic acid is an oxidation product of glucose in animal tissue (Salmony and Whitehead, 1954). This metabolite has complex taste (acid, bitter, metallic). Gluconic acid was negatively correlated to pyruvic acid, 3-methylheptane and (E)-2-octene, with the two latter compounds positively correlated to grassflavor. The acid was found in all animals and can possibly be a universal marker of grass flavor.

Volatiles, formic acid, β-carophyllene, 3-methylphenol, 2-ethylfuran, were positively correlated to grassflavor. 3-methylphenol was above threshold in this group and actually had aflavor that makes it likely to influence grassflavor (see Table S-1), but not as a universal marker. In addition, four volatiles were identified as negatively af- fecting grassflavor. Propanal (r = -0.39,P< 0.001) and 2-propenal (r

= -0.33,P= 0.002) were negatively correlated with grassflavor, although both compounds were above threshold and described as pun- gent. (E)-2-nonenal and 2-heptadecanone also showed negative correlation (r = -0.33, P= 0.024) with grass flavor. (E)-2-nonenal was above threshold and has aflavor that would normally be associated with lipid oxidation (Kosowska et al., 2017). Although the trained panelists could not identify rancidity as a discriminator this cannot ex- clude lipid oxidation as an undesirable process that ultimately leads to development of off-flavors.

Group of antioxidants composed of 4-hydroxyproline, ascorbic acid, galactitol and uridine where negatively correlated with grassflavor. 4- hydroxyproline as a sweet taste compound was identified in stewed beef juice (Schlichtherle-Cerny and Grosch, 1998). Sweet taste properties of 4-hydroxyproline and antioxidant function of ascorbic acid (Howes et al., 2015) could suppress grassyflavor. Antioxidant compounds were below thresholds and therefore had no direct influence on flavor. In addition, antioxidant capacity of muscle is associated with vitamin E as a fat soluble vitamin (Howes et al., 2015).Hopkins et al.

(2013)found that vitamin E can prevent lipid oxidation, when PUFA was present at high levels, i.e. at concentration above 2.95 mg/kg of muscle. The old sheep studied here had high content of vitamin E (2.5 mg/kg;Bjelanovićet al., 2015), close to reported threshold. This can possibly explain the relatively low number of lipid-oxidation products that correlated with grassﬂavor and the absence of rancidity.

Glucose was above threshold and its sweetness suppressed grass ﬂavor. This sugar had negative correlation to grassﬂavor (r = -0.50,

P< 0.001). In addition, glucose is metabolically related to many compounds; positive correlation to fructose, galactose and mannose-6- phosphate (r = 0.83, 0.84, 0.50, respectively,P< 0.001) was found.

Sugars seems to play a significant role in suppression of off-flavors in lamb/sheep and this is most obvious for gamy and grassflavor. Myo- inositol possibly has the same role as suggested for the gamyflavor.

3.3.3. Bitterﬂavor

Bitterflavor was the third best explainedflavor attribute (27%) that did not show high correlation to gamy or grass flavor. Maximum 6 clusters were suggested by PLS analysis. If variables with no correlation to bitterflavor were removed, i.e. the data set of volatiles was reduced, 40% of the variation in bitterflavor was explained.Fig. 1c shows positive correlation of hypotaurine (r = 0.52, P< 0.001) with bitter flavor. Although, the threshold concentration for hypotaurine was by the criteria used, higher than identified concentration in lamb and sheep, participants recognized bitter flavor at the lowest tested concentration (6.25 mg/kg). Thus, since the lowest tested concentration was close to highest identified concentration in some animals (4.7 mg/

kg) it is plausible that hypotaurine contributed to bitterness in sheep.

The metabolites 4-hydroxyproline, hypotaurine, O-phosphoryl- ethanolamine, were positively correlated with bitterflavor, while galactonic acid showed a negative correlation (r = -0.46, P< 0.001) with this attribute. Additionally, galactonic acid positively correlated with several sugar phosphates (e.g. glucose-6-phosphate, r = 0.53, P< 0.001), known as sweet compounds, and that may explain its negative contribution to bitter flavor. O-phosphoryl-ethanolamine, positively correlated (r = 0.36,P< 0.05) with bitterflavor, having a direct or indirect effect.Mabuchi et al. (2018)reported positive correlation of O-phosphoryl-ethanolamine with umami taste and hypotaurine with acidic bitterness taste offish.

Glyceric acid and galactitol showed positive correlation (r = 0.31, P= 0.003) to bitter ﬂavor. Glyceric acid positively correlated with gluconic acid and pyruvic acid (r = 0.60 and r = 0.54, P< 0.001, respectively). Galactitol was below threshold and positively correlated with glucuronic acid (r = 0.73) together with cysteine (r = 0.54), arabitol (r = 0.53) and inosine (r = 0.55), all withP< 0.001. Cysteine was above threshold in all samples and showed positive correlation (r

= 0.15,P= 0.004) to bitterness, without being picked out directly as important for describing bitterflavor. Negative correlation (r = -0.28, P= 0.01) of nicotinic acid with bitterflavor was surprising since this compound has a bitterflavor. Possible explanation for this phenomenon is the presence of nicotinic acid in the concentration far belowflavor threshold. In addition, positive correlation of nicotinic acid to glycine (r

= 0.49,P< 0.001), as slightly sweet compound, modulated bitter perception of this metabolite.

Other positive correlation was observed for hexane with hypotaurine (r = 0.39,P< 0.05) and bitterﬂavor (r = 0.36,P< 0.05).

However, the contribution of hexane to bitterflavor in this study was small since this volatile compound was present in concentration below flavor threshold. In addition, (E)-2-pentanal showed positive correlation to bitterflavor (r = 0.51, P< 0.05) and it was present in concentration aboveflavor threshold, but could not directly contribute to bitterflavor as it was described with fruity/greenflavor. Three volatile compounds, heptanal, (E,Z)-2,4-heptadienal (r = -0.34 and r = -0.36, respectively,P< 0.05) and 2-undecanone (r = -0.34,P< 0.01), were above threshold level and negatively correlated with bitterflavor. Fatty perception of these compounds possibly had negative effect on bitter attribute.

3.3.4. Metallicﬂavor

Metallicflavor, including taste and olfactory sensation, may results from iron compounds (Mitterer-Daltoé et al., 2012). In our study metallicflavor was the 4^thbest explainedflavor attribute (19% explained, validated). There were max 3 independent factors. The issue with metallic flavor was that it correlated significantly to bitter (R-

(7)

square = 0.63,P< 0.001, not validated). This was apparent from the compound listed below. There were no really new compounds to explain metallicflavor but arabitol was selected as indirectly involved in bitter flavor. By selecting a subset of compounds it was possible to explain 30% of the variation in metallicflavor; this means that it was not a well explained attribute. Among volatiles hexane may have a small positive (r = 0.33,P= 0.002) influence on metallicflavor. Most lipid volatiles had negative effects on metallicflavor.

Acidicﬂavor was not explained when model validation was used.

The nominally lowest P value was to galactonic acid (r = 0.1, P= 0.37).

3.4. Relationship between identifiedflavor compounds and meat origin Tracing the origin of products is important for authentication of meat from different production systems. Therefore additional explora- tion of marker of origin was therefore carried out.

Principal component analysis (PCA) was used to diﬀerentiate the animal population based on identiﬁed volatile compounds isolated from adipose tissues and metabolites from lean meat of lamb and sheep.

PCA, the first two components, carried out on the volatile compounds isolated from lamb, young and old sheep from BH, MN, and NW, are shown in Fig. 2a. A clear differentiation between volatile profiles of 4y old sheep that belonged to two phenotypes of the

Pramenka breed was observed. MN 4y sheep volatile profile was determined by 14 compounds, including gamy and grass-related (2-methylheptane, 3-methylheptane, (E)-2-octene and 3-methyphenol)flavor compounds. The meat from MN 4y was also significantly more gamy (P< 0.05) and nominally more grassy than meat from BH 4y and NW 4 y. Butanol and 3-methylphenol (latter clearly above threshold) were only identified in MN animals presenting potential biomarkers of this production system. Furthermore,β-caryophyllene, almost exclusively synthesized in plant tissue, was identified only in BH 4y sheep. The role of sesquiterpenes in lamb and sheepflavor profiles needs further investigation regarding seasonal changes.

Four animal groups (BH lamb, NW lamb, NW 2y sheep, NW 4y sheep) that showed poor separation infirst PCA, was used to develop a second PCA model.Fig. 2b shows that the volatile profile of BH lamb did not clustered together with NW animals. Two volatiles were associated with BH lamb profile, among them (E)-2-pentenal which has been proposed as a potential, indirect biomarker of bitterflavor in our data. BH lambs had the most bitter tasting meat and was significantly (P< 0.05) more bitter than all other meats.

In order to obtain more information about relationship between metabolites and animal population, a third PCA analysis were performed. The PCA plot inFig. 2c showed a clear diﬀerentiation for all animal groups, and the animals from the same production system clustered together. Although NW animals clustered, no characteristic

Fig. 2.Diﬀerentiation of animal groups (P < 0.001) based on: (a) volatile compounds isolated from heated adipose tissue of all animal groups (BH lamb, BH 4y, MN 4y, NW lamb, NW 2y, and NW 4y); (b) volatile compounds isolated from heated adipose tissue of four animal groups (BH lamb, NW lamb, NW 2y and NW 4y) that showed poor separation when all animal groups were included in PCA; (c) metabolite compounds isolated from lean meat of all animal groups.

(8)

metabolite was identiﬁed. Gluconic and pyruvic acid related to MN animals’meat and supports the gamy and grassﬂavor note of the meat.

BH 4y meat pattern was defined by high concentration of essential amino acids that clustered with bitter flavor and these amino acids seem to support bitterness in the more bitter meats; NW lambs and NW 4y being the least bitter meats. These two meats were also the least metallic. Furthermore, antioxidant compounds that suppressed off- flavor properties may contribute to the unique BH lamb metabolite pattern.

Our results indicate that PCA plots offer an interesting approach regarding discrimination of animals from different production system using flavor markers. However, some of the markers might change depending of pasture season.

Declaration of Competing Interest

Per Berg is employed in the Research Unit of the cooperative meat production in Norway. He has provided the Norwegian animals, but not inﬂuenced the data treatment and conclusions. All authors declare no competingﬁnancial interest.

Acknowledgements

The work was funded by the Research Council of Norway (Grants FR184846/I10and225309) and by the Norwegian Ministry for Foreign Aﬀairs (Grant 19028), while a PhD scholarship was provided by the Norwegian State Educational Loan fund (“Lånekassen”).

The staffat Nortura Gol, "BB" Kotor Varoš, and "Franca" Bijelo Polje are thanked for their assistance at the slaughter line. The author from Nortura SA contributed to the planning phase by organizing the collection of samples and also assisted during the manuscript writing, but had no influence on the choice of methodology, registration of data, choice of statistical methods and interpretation of results. We would like to thank Kari Olsen for the technical help using HS-GC/MS analysis, Božidarka Marković, Goran Vučić and Božo Važić for their help in sample collection. Erik Slinde is thanked for helping out with metabolite pathway identification.

Appendix A. Supplementary data

Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.smallrumres.2019.09.

022.

References

Almela, E., Jordán, M.J., Martínez, C., Sotomayor, J.A., Bedia, M., Bañón, S., 2010. Ewe’s diet (pasture vs grain-based feed) aﬀects volatile proﬁle of cooked meat from light lamb. J. Agric. Food Chem. 58, 9641–9646.

AMSA, 1995. Research Guidelines for Cookery, Sensory Evaluation and Instrumental Tenderness Measurements of Fresh Meat. American Meat Science Association, National Live Stock and Meat Board., Chicago, Illinois, USA.

Bjelanović, M., Grabež, V., Vučić, G., Martinović, A., Lima, L.R., Marković, B., Egelandsdal, B., 2015. Eﬀect of diﬀerent production system on carcass and meat quality of sheep and lamb from Western Balkan and Norway. J. Biotech. Anim. Husb.

31, 203–221.

Bueno, M., Mar Campo, M., Cacho, J., Ferreira, V., Escudero, A., 2014. A model ex- plaining and predicting lambﬂavour from the aroma-active chemical compounds released upon grilling light lamb loins. Meat Sci. 98, 622–628.

Cadwallader, K.R., Singh, T.K., 2009. Flavours and oﬀ-ﬂavours in milk and dairy products. In: In: McSweeney, P., Fox, P.F. (Eds.), Advanced Dairy Chemistry Vol 3.

Springer, New York, pp. 631–690.

Campo, M.M., Nute, G.R., Wood, J.D., Elmore, S.J., Mottram, D.S., Enser, M., 2003.

Modelling the eﬀect of fatty acids in odour development of cooked meat in vitro: part I—sensory perception. Meat Sci. 63 (3), 367–375.

Caporaso, F., Sink, J.D., Dimick, P.S., Mussinan, C.J., Sanderson, A., 1977. Volatileﬂavor constituents of ovine adipose tissue. J. Agric. Food Chem. 25 (6), 1230–1234.

Curry, J.W., Hohl, R., Noakes, T.D., Kohn, T.A., 2012. High oxidative capacity and type IIxﬁbre content in springbok and fallow deer skeletal muscle suggest fast sprinters with a resistance to fatigue. J. Exp. Biol. 215 (22), 3997–4005.

Drauz, K., Grayson, I., Kleemann, A., Krimmer, H.P., Leuchtenberger, W., Weckbecker, C.,

2007. Amino Acids. Ullmann’s Encyclopedia of Industrial Chemistry.

Dunkel, A., Hofmann, T., 2009. Sensory-directed identiﬁcation ofβ-alanyl dipeptides as contributors to the thick-sour and white-meaty orosensation induced by chicken broth. J. Agric. Food Chem. 57, 9867–9877.

Elmore, J.S., Mottram, D.S., Enser, M., Wood, J.D., 2000. The eﬀects of diet and breed on the volatile compounds of cooked lamb. Meat Sci. 55, 149–159.

Elmore, J.S., Cooper, S.L., Enser, M., Mottram, D.S., Sinclair, L.A., Wilkinson, R.G., Wood, J.D., 2005. Dietary manipulation of fatty acid composition in lamb meat and its eﬀect on the volatile aroma compounds of grilled lamb. Meat Sci. 69, 233–242.

Gomez, F.E., Cassís-Nosthas, L., Morales-de-León, J.C., Bourges, H., 2004. Detection and recognition thresholds to the 4 basic tastes in Mexican patients with primary Sjögren’s syndrome. Eur. J. Clin. Nutr. 58, 629–636.

Hopkins, D., Lamb, T., Kerr, M., van de Ven, R., Ponnampalam, E., 2013. Examination of the eﬀect of ageing and temperature at rigor on color stability of lamb meat. Meat Sci.

95 (2), 311–316.

Hornstein, I., Crowe, P.F., 1963. Foodﬂavors and odors, meatﬂavor: lamb. J. Agric. Food Chem. 11 (2), 147–149.

Howes, N.L., Bekhit, A.E.A., Burritt, D.J., Campbell, A.W., 2015. Opportunities and im- plications of pasture-based lamb fattening to enhance the long-chain fatty acid composition in meat. Comp. Rev. Food Sci: Food Safety. 14, 22–36.

Hui, Y.H., Evranuz, E.Ö., 2012. Handbook of Fermented Food and Beverage Technology, second ed. CRS Press.

Kosowska, M., Majcher, M.A., Fortuna, T., 2017. Volatile compounds in meat and meat products. Food Sci. Technol, Campinas. 37 (1), 1–7.

Koutsidis, G., Elmore, J.S., Oruna-Concha, M.J., Campo, M.M., Wood, J.D., Mottram, D.S., 2008. Water-soluble precursors of beefﬂavour. Part II: Eﬀect of post-mortem con- ditioning. Meat Sci. 79 (2), 270–277.

Lawless, H.T., Heymann, H., 2010. Sensory Evaluation of Food, Principles and Practices, second edition. Springer.

Lind, V., Berg, J., Eilertsen, S.M., Hersleth, M., Eik, L.O., 2011. Eﬀect of gender on meat quality in lamb from extensive and intensive grazing systems when slaughtered at the end of the growing season. Meat Sci. 88, 305–310.

Lugaz, O., Pillias, A.-M., Boireau-Ducept, N., Faurion, A., 2005. Time–intensity evaluation of acid taste in subjects with saliva highﬂow and lowﬂow rates for acids of various chemical properties. Chem. Senses 30 (1), 89–103.

Mabuchi, R., Ishimaru, A., Tanaka, M., Kawaguchi, O., Tanimoto, S., 2018. Metabolic proﬁling ofﬁsh meat by GC-MS analysis, and correlations with taste attributes obtained using an electronic tongue. Metabolites 9 (1), 1.

Madruga, M., Dantas, I., Queiroz, A., Brasil, L., Ishihara, Y., 2013. Volatiles and water- and fat soluble precursors of Saanen goat and cross Suﬀolk lambﬂavour. Molecules 18, 2150–2165.

Maruri, J.L., Larick, D.K., 1992. Volatile concentration andﬂavor of beef as inﬂuenced by diet. J. Food Sci. 57 (6), 1275–1281.

Meinert, L., Schäfer, A., Bjergegaard, C., Aaslyng, M.D., Bredie, W.L.P., 2009. Comparison of glucose, glucose-6-phosphate, ribose, and mannose asﬂavour precursors in pork;

the eﬀect of monosaccharide addition onﬂavour generation. Meat Sci. 81 (3), 419–425.

Mitterer-Daltoé, M.L., Treptow, R.O., Martins, E., Martins, V.M.V., Queiroz, M.I., 2012.

Selecting and training a panel to evaluate the metallic sensation of meat. Food Sci.

Technol. Res. 18 (2), 279–286.

Nishimura, T., Rhue, M.R., Okitani, A., Kato, H., 1988. Components contributing to the improvement of meat taste during storage. Agric. Biol. Chem. 52 (9), 2323–2330.

O’Fallon, J.V., Busboom, J.R., Nelson, M.L., Gaskins, C.T., 2007. A direct method for fatty acid methyl ester synthesis: application to wet meat tissues, oils, and feedstuﬀs. J.

Anim. Sci. 85 (6), 1511–1521.

Osorio, M.T., Zumalacárregui, H.M., Cabeza, E.A., Figueira, A., Mateo, J., 2008. Eﬀect of rearing system on some meat quality traits and volatile compounds of suckling lamb meat. Small Rumin. Res. 78 (1–3), 1–12.

Perfumer and Flavorist, 2017a. Dimethyl Sulfone. Accessed 2017-07-04, from. http://

www.thegoodscentscompany.com/data/rw1260371.html.

Perfumer and Flavorist, 2017b. 2-heptadecanone. Accessed 2017-07-04, from. http://

www.thegoodscentscompany.com/data/rw1436641.html.

Perfumer and Flavorist, 2017c. Hypotaurine. Accessed 2017-07-08, from. http://www.

thegoodscentscompany.com/data/rw1684611.html.

Resconi, V.C., Mar Campo, M., Montossi, F., Ferreira, V., Sañudo, C., Escudero, A., 2010.

Relationship between odour-active compounds andﬂavour perception in meat from lambs fed diﬀerent diets. Meat Sci. 85 (4), 700–706.

Ronningen, I., 2016. Untargeted Flavoromics to Identify Flavor Active Compounds.

University of Minnesota Ph.D. dissertation, pp. 188.

Salmony, D., Whitehead, J.K., 1954. Studies on the oxidation of gluconate by animal tissues. Biochem. J. 58 (3), 408–413.

Schlichtherle-Cerny, H., Grosch, W.Z., 1998. Evaluation of taste compounds of stewed beef juice. Z. Lebensm.-Unters. -Forsch., A Food Res. Technol. (Print) 207, 369–376.

Schreurs, N.M., Lane, G.A., Tavendale, M.H., Barry, T.N., McNabb, W.C., 2008. Pastoral ﬂavor in meat products from ruminants fed fresh forages and its amelioration by forage condensed tannins. Anim. Feed Sci. Technol. 193–221.

Sink, J.D., Caporaso, F., 1977. Lamb and muttonﬂavour: contributing factors and chemical aspects. Meat Sci. 1, 119–127.

Sissener, N.H., Hemre, G.I., Lall, S.P., Sagstad, A., Petersen, K., Williams, J., Rohloﬀ, J., Sanden, M., 2011. Are apparent negative eﬀects of feeding GM MON810 maize to Atlantic salmon, Salmo salar, caused by confounding factors? Br. J. Nutr. 106, 42–56.

Sivadier, G., Ratel, J., Bouvier, F., Engel, E., 2008. Authentication of meat products:

determination of animal feeding by parallel GC-MS analysis of three adipose tissues.

J. Agric. Food Chem. 56 (21), 9803–9812.

Torrico, D.D., Sae-Eaw, A., Sriwattana, S., Boeneke, C., Prinyawiwatkul, W., 2015. Oil-in- water emulsion exhibits bitterness-suppressing eﬀects in a sensory threshold study. J.

(9)

Food Sci. 80, 1404–1411.

Umano, K., Shibamoto, T., 1987. Analysis of headspace volatiles from overheated beef fat.

J. Agric. Food Chem. 35 (1), 14–18.

Van Ba, H., Hwang, I., Jeong, D., Touseef, A., 2012. Principle of Meat Aroma Flavors and Future Prospect. Chapter 7. InTech, pp. 145–176.

Vasta, V., Ratel, J., Engel, E., 2007. Mass spectrometry analysis of volatile compounds in raw meat for the authentication of the feeding background of farm animals. J. Agri.

Food Chem. 55 (12), 4630–4639.

Vasta, V., Ventura, V., Luciano, G., Andronico, V., Pagano, R.I., Scerra, M., Biondi, L., Avondo, M., Priolo, A., 2012. The volatile compounds in lamb fat are aﬀected by the time of grazing. Meat Sci. 90 (2), 451–456.

Volden, J., Bjelanovic, M., Vogt, G., Slinde, E., Skaugen, M., Nordvi, B., Egelandsdal, B., 2011. Oxidation progress in an emulsion made from metmyoglobin and diﬀerent triacylglycerols. Food Chem. 128, 854–863.

Warner, R.D., Ferguson, D.M., Cottrell, J.J., Knee, B.W., 2007. Acute stress induced by the preslaughter use of electric prodders causes tougher beef meat. Aust. J. Exp. Agric.

47, 782–788.

Watkins, P.J., Frank, D., Singh, T.K., Young, O.A., Warner, R.D., 2013. Sheep meatflavour and the effect of different feeding systems: A Review. J. Agric.Food Chem. 61,

3561–3579.

Watkins, P.J., Rose, G., Salvatore, L., Allen, D., Tucman, D., Warner, R.D., Dunshea, F.R., Pethick, D.W., 2010. Age and nutrition inﬂuence the concentrations of three branched chain fatty acids in sheep fat from Australian abattoirs. Meat Sci. 86, 594–599.

Wieser, H., Jugel, H., Belitz, H.D., 1977. Zusammenhänge zwischen struktur und Süßgeschmack Bei Aminosäuren. Z. Lebensm. Unters. Forch. 164, 277.

Wong, E., 1975. Muttonﬂavour. Food Tech. N. Z. 13–15.

Wong, E., Johnson, C.B., Nixon, L.N., 1975a. The contribution of 4-methylocanoic (hir- cinoic) acid to mutton and goat meatﬂavour. N. Z. J. Agric. Res. 18, 261–266.

Wong, E., Nixon, L.N., Johnson, C.B., 1975b. Volatile medium chain fatty acids and muttonﬂavor. J. Agric. Food Chem. 23, 495–498.

Wood, J.D., Enser, M., Fisher, A.V., Nute, G.R., Richardson, R.I., Sheard, P.R., 1999.

Manipulating meat quality and composition. Proc. Nutr. Soc. India 58, 363–370.

Yi, G., Haug, A., Nyquist, N.F., Egelandsdal, B., 2013. Hydroperoxide formation in different lean meats. Food Chem. 141 (3), 2656–2665.

Young, A., Berdagué, J.-L., Viallon, C., Rousset-Akrim, S., Theriez, M., 1997. Fat-borne volatiles and sheepmeat odour. Meat Sci. 45 (2), 183–200.

Young, O.A., Lane, G.A., Priolo, A., Fraser, K., 2003. Pastoral and speciesﬂavor in lambs raised on pasture, lucerne or maize. J. Sci. Food Agric. 83, 94–104.